[Bug middle-end/109370] New: Missed optimization for std::optional branchless unwrapping

[Bug middle-end/109370] New: Missed optimization for std::optional branchless unwrapping

[andre.schackier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109370

            Bug ID: 109370
           Summary: Missed optimization for std::optional branchless
                    unwrapping
           Product: gcc
           Version: 12.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: andre.schackier at gmail dot com
  Target Milestone: ---

Given the following source code godbolt: https://godbolt.org/z/vW6ebqafK

#include <optional>

int f(std::optional<int>&& o) {
    if (!o) return -1;

    return *o;
}

gcc with '-O3' generates:
f(std::optional<int>&&):
        cmp     BYTE PTR [rdi+4], 0
        je      .L3
        mov     eax, DWORD PTR [rdi]
        ret
.L3:
        mov     eax, -1
        ret

while clang generates:
f(std::optional<int>&&):                    # @f(std::optional<int>&&)
        xor     eax, eax
        cmp     byte ptr [rdi + 4], 1
        sbb     eax, eax
        or      eax, dword ptr [rdi]
        ret

[Bug middle-end/109370] Missed optimization for std::optional branchless unwrapping

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109370

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement
           Keywords|                            |missed-optimization

[Bug middle-end/109370] Missed optimization for std::optional branchless unwrapping

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109370

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I am not so sure this is worse.

Also I find LLVM code generataion depdedent on if the argument is a pointer vs
a reference (rvalue or normal):
```
struct a
{
        int b;
        bool c;
};

int f(struct a  &o) {
    if (!o.c) return -1;
    return o.b;
}
int f(struct a  *o) {
    if (!o->c) return -1;
    return o->b;
}
```

[Bug rtl-optimization/109370] Missed optimization for std::optional branchless unwrapping

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109370

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|middle-end                  |rtl-optimization

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note we would need to know that dereferencing 'o' is OK (not sure if RTL knows
this), and this if-conversion is certainly RTL territorry because of costing. 
On x86 I'd expect the branchy version to be faster.